What Are Donchian Channels?
Donchian Channels plot the highest high and the lowest low over a lookback period, usually 20 bars. With the offset at zero the channel redraws to include every new extreme, so price can never close outside it; an offset of one bar is what makes a breakout possible.
Richard Donchian’s channel is the simplest thing in this glossary and it has a default setting that makes it useless. Both facts are worth knowing.
How it works
Two lines and no arithmetic.
upper = highest high of the last 20 bars
lower = lowest low of the last 20 bars
No averaging, no weighting, no smoothing. That is why the lines are stepped rather than curved — they only move when a new extreme enters or an old one leaves the window.
The default cannot be broken
This is the fact the page exists for.
With the offset at zero, the upper line is the highest high including the current bar. So the moment price makes a new high, the line moves up to meet it.
Measured on this chart: 0 closes outside the channel. Not “few” — none, and there could not be any, because the definition includes the bar you are testing against it.
Shift the line one bar back and it becomes a level. Now the upper line is the highest high of the previous 20 bars, which today’s price can genuinely exceed.
Same data, same length: 3 breaks instead of 0.
Most platforms default the offset to zero, which produces a beautiful chart where every rally hugs the top of the channel and nothing is ever tradeable.
What it is really for
It is a breakout rule with a line attached.
The best-known use is the Turtle system of the early 1980s: buy a 20-day high, exit on a 10-day low, with a slower 55-day version running alongside. The channel is not analysis — it is the rule made visible.
Which is a genuine virtue. There is no judgment in it, so two people looking at the same chart with the same settings see the same level, and neither can move it after the fact — the failure the trend lines page spends its length on.
The middle line
It is (upper + lower) / 2 — halfway between two extremes, which is not the same thing as a
moving average of the closes.
Everything that happened between those two extremes is invisible to it. A market that spent nineteen bars near its low and one near its high produces the same midline as one that split its time evenly.
That makes it a poor trend line and a reasonable trailing exit, which is roughly how the Turtles used their shorter channel.
The settings
Two numbers: the lookback and the offset.
The offset is the one that matters and it is covered above: zero makes the indicator decorative, one makes it a level.
The lookback decides how often a break happens. Measured here: 3 breaks at 20 bars and 4 at 10. A shorter window means more highs qualify as new, which is more signals from the same market.
A worked example
Set the offset to 1 before anything else. If your platform does not offer an offset, the channel you are looking at cannot be broken and is not a level.
The upper line is a price you could have written down 20 bars ago. That is its whole value — it is fixed, public, and identical on every screen running the same settings.
A close above it is the breakout, with the same caveats the breakout page covers: a wick through is not a break, and most breaks fail.
The exit is a shorter channel in the other direction — the Turtle construction, and a trailing stop with no judgment in it.
The original data
Across our study of 24,971 trading videos, 61 cover Donchian Channels. The median one gets 8,797 views, 85% never pass 50,000, and the median length is 8.1 minutes.
That median is well above Keltner Channels at 2,967 on a field half the size, which is a little surprising for the more obscure of the two.
The corpus carries description text for 54 of those 61, and across those 54, one mentions invalidation, failure, or what a bad read looks like.
Not one of them mentions the offset, which is the setting that decides whether the indicator does anything at all.
When it fails
Sideways, every edge is a break
In a trading range the channel narrows to the range itself, and each push at either edge registers as a new extreme.
That is a breakout system in the condition breakout systems lose in, which is the honest limit of the whole approach.
The lines only move on extremes
A channel does not know that the last nineteen bars were quiet. It reports the one violent bar from twenty sessions ago until that bar drops out of the window, at which point the line jumps.
Those jumps are not events. They are old data expiring, and they can look like a sudden tightening that means nothing.
It has no volatility in it
Unlike Keltner or Bollinger Bands, width here is just the recent range. In a quiet market the channel is tight and a trivial move breaks it, and nothing in the tool adjusts for that.
The Turtles knew this and fixed it outside the indicator. Their position size came from average true range (ATR), so a break of a narrow channel bought a larger position than a break of a wide one, and the money at risk stayed constant either way.
That is the pattern worth taking from the whole system: the channel decides whether, and a separate volatility measure decides how much. Using the channel for both is asking one very simple tool to do two jobs, and it can only do the first.
You looked once the break had run
Price at the top of the channel is where every real breakout and every failed one both begin.
Related
Breakout is what a Donchian break actually is, including how often it fails.
Trading range is the condition that turns the channel into a signal generator with nothing behind it.
And Keltner Channels is the same shape of tool built from volatility instead of extremes.
The offset thing is the single most useful five minutes I have spent inside an indicator settings panel. I had this on a chart for months wondering why the breakouts always looked perfect in hindsight and never seemed tradeable, and the answer was that the line was moving to wherever price went. Once I set the offset to one it started disagreeing with price, which is the only way a level can be worth anything.
— Michael Whitman, from this video
This page is educational, not financial advice. Test every idea on your own charts before risking money.